High Availability Motion Sensor with Nonlinear Interferometry and AI

Self-mixing interferometry is a well-established technique in which a laser (often a semiconductor laser) is submitted to an optical feedback after reflection of the beam on a target. The operating point of the laser is altered depending on the phase of the reflected field. Among many applications, this method can be used to measure the displacement of a target along the propagation axis [1]. One can monitor these alterations of the operation point by a direct measurement of the voltage at the terminals of the laser diode. When the target moves in one direction or an other, the signal follows a sawtooth shape, for which each transition corresponds to a
$\lambda/2$
variation of the feedback phase. In general, reconstructing the displacement from an interferometric signal is made by identifying and counting peaks, or fitting a model on the signal. However, the actual amount of light fed back into the laser has a large impact on the shape of the signal and (in the case of non-cooperative targets), this amount of feedback is hard to control. This particularity makes the technique harder to use with non-cooperative targets, in which speckle phenomena can easily bring the system outside of the desired parameter range. Thus, many approaches have been taken to control this feedback level and/or to design ad-hoc signal processing techniques for the reconstruction of the displacement from the interferometric signal [2].